Network medicine framework shows that proximity of polyphenol targets and disease proteins predicts therapeutic effects of polyphenols - Barabasi
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Articles
https://doi.org/10.1038/s43016-021-00243-7
Network medicine framework shows that proximity
of polyphenol targets and disease proteins predicts
therapeutic effects of polyphenols
Italo F. do Valle1, Harvey G. Roweth2,3, Michael W. Malloy2,3, Sofia Moco 4
, Denis Barron4,
1,6,7 ✉
Elisabeth Battinelli2,3, Joseph Loscalzo 3,5 and Albert-László Barabási
Polyphenols, natural products present in plant-based foods, play a protective role against several complex diseases through
their antioxidant activity and by diverse molecular mechanisms. Here we develop a network medicine framework to uncover
mechanisms for the effects of polyphenols on health by considering the molecular interactions between polyphenol protein
targets and proteins associated with diseases. We find that the protein targets of polyphenols cluster in specific neighbour-
hoods of the human interactome, whose network proximity to disease proteins is predictive of the molecule’s known thera-
peutic effects. The methodology recovers known associations, such as the effect of epigallocatechin-3-O-gallate on type 2
diabetes, and predicts that rosmarinic acid has a direct impact on platelet function, representing a novel mechanism through
which it could affect cardiovascular health. We experimentally confirm that rosmarinic acid inhibits platelet aggregation and
α-granule secretion through inhibition of protein tyrosine phosphorylation, offering direct support for the predicted molecu-
lar mechanism. Our framework represents a starting point for mechanistic interpretation of the health effects underlying
food-related compounds, allowing us to integrate into a predictive framework knowledge on food metabolism, bioavailability
and drug interaction.
D
iet plays a defining role in human health. Indeed, while conflicting conclusions about the beneficial effects of resveratrol on
poor diet can substantially increase the risk for coronary glycemic control in T2D patients12,13. We therefore need a frame-
heart disease and type 2 diabetes mellitus (T2D), a healthy work to interpret the evidence present in the literature and to offer
diet can play a protective role, even mitigating genetic risk for in-depth mechanistic predictions of the molecular pathways respon-
coronary heart disease1. Polyphenols are a class of compounds sible for the health implications of polyphenols present in the diet.
present in plant-based foods, including fruits, vegetables, nuts, Ultimately, these insights could help us provide evidence on causal
seeds, beans (for example coffee and cocoa), herbs, spices, tea and diet–health associations as well as guidelines of food consumption
wine, that play a well-documented protective role as antioxidants, for different individuals and help to develop novel diagnostic and
which affect several diseases, from cancer to T2D, cardiovascular therapeutic strategies that could lead to the synthesis of novel drugs.
and neurodegenerative diseases2,3. Previous efforts have profiled Here, we address this challenge by developing a network medi-
over 500 polyphenols in more than 400 foods4,5 and have docu- cine framework to capture the molecular interactions between
mented the high diversity of polyphenols to which humans are polyphenols and their cellular binding targets, unveiling their rela-
exposed through their diet, ranging from flavonoids to phenolic tionship to complex diseases. The developed framework is based on
acids, lignans and stilbenes. the human interactome, a comprehensive subcellular network con-
The underlying molecular mechanisms through which specific sisting of all known physical interactions between human proteins
polyphenols exert their beneficial effects on human health remain that has been validated previously as a platform for understand-
largely unexplored. From a mechanistic perspective, dietary poly- ing disease mechanisms14,15, rational drug target identification and
phenols are not engaged in endogenous metabolic processes of drug repurposing16,17.
anabolism and catabolism, but rather affect human health through We find that the proteins to which polyphenols bind form identifi-
their anti- or pro-oxidant activity6 by binding to proteins and modu- able neighbourhoods in the human interactome, allowing us to dem-
lating their activity7,8, interacting with digestive enzymes9 and mod- onstrate that the proximity between polyphenol targets and proteins
ulating gut microbiota growth10,11. Yet the variety of experimental associated with specific diseases is predictive of the known therapeu-
settings and the limited scope of studies that explore the molecular tic effects of polyphenols. Finally, we unveil the potential therapeutic
effects of polyphenols have, to date, offered a range of often conflict- effects of rosmarinic acid (RA) on vascular diseases (VD), predict-
ing evidence. For example, two clinical trials, both limited in terms ing that its mechanism of action is related to modulation of platelet
of the number of subjects and the intervention periods, resulted in function. We confirm this prediction by experiments that indicate
Network Science Institute and Department of Physics, Northeastern University, Boston, MA, USA. 2Division of Hematology, Department of Medicine,
1
Brigham and Women’s Hospital, Boston, MA, USA. 3Harvard Medical School, Boston, MA, USA. 4Nestlé Institute of Health Sciences, Lausanne,
Switzerland. 5Department of Medicine, Brigham and Women’s Hospital, Boston, MA, USA. 6Channing Division of Network Medicine, Department of
Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA. 7Department of Network and Data Science, Central European
University, Budapest, Hungary. ✉e-mail: barabasi@gmail.com
Nature Food | VOL 2 | March 2021 | 143–155 | www.nature.com/natfood 143Articles NaTure FOOd
that RA modulates platelet function in vitro by inhibiting tyrosine and physiological and biochemical studies have shown that EGCG
protein phosphorylation. Altogether, our results demonstrate that presents glucose-lowering effects in both in vitro and in vivo mod-
the network-based relationship between disease proteins and poly- els22,23. We identified 54 experimentally validated EGCG protein
phenol targets offers a tool to systematically unveil the health effects targets and mapped them to the interactome, finding that the
of polyphenols. EGCG targets form an LCC of 17 proteins (Z-score = 7.61) (Fig. 3a).
We also computed the network-based distance between EGCG tar-
Results gets and 83 proteins associated with T2D, finding that the two sets
Polyphenol targets cluster in specific functional neighbour- are significantly proximal to each other. We ranked all 299 diseases
hoods of the interactome. We mapped the targets of 65 polyphe- based on their network proximity to the EGCG targets to deter-
nols (Methods) to the human interactome, consisting of 17,651 mine whether we could recover the 82 diseases in which EGCG has
proteins and 351,393 interactions (Fig. 1a,b). We find that 19 of known therapeutic effects according to the comparative toxicoge-
the 65 polyphenols have only one protein target, while a few poly- nomics database (CTD)24. By this analysis, we were able to recover
phenols have an exceptionally large number of targets (Fig. 1c). 15 previously known therapeutic associations among the top 20
We computed the Jaccard index (JI) of the protein targets of each ranked diseases (Table 1), confirming that network proximity can
polyphenol pair, finding only a limited similarity of targets among discriminate between known and unknown disease associations for
different polyphenols (average JI = 0.0206) (Supplementary Fig. 1a). polyphenols, as previously confirmed for drugs16,17.
Even though the average JI is small, it is still significantly higher We expanded these methods to all polyphenol–disease pairs to
(Z = 147, Supplementary Fig. 1b) than the JI expected if the poly- predict diseases for which specific polyphenols might have thera-
phenol targets were randomly assigned from the pool of all network peutic effects. For this analysis, we grouped all 19,435 polyphenol–
proteins with degrees matching the original set. This finding sug- disease associations between 65 polyphenols and 299 diseases into
gests that while each polyphenol targets a specific set of proteins, known (1,525) and unknown (17,910) associations. The known
their targets are confined to a common pool of proteins, likely deter- polyphenol–disease set was retrieved from the CTD, which is limited
mined by commonalities in the polyphenol-binding domains of the to manually curated associations for which there is literature-based
three-dimensional structure of the protein targets18. Gene ontology evidence. For each polyphenol, we tested how well network prox-
enrichment analysis recovers existing mechanisms8 and also helps imity discriminates between the known and unknown sets by eval-
identify new processes related to polyphenol protein targets, such uating the area under the curve (AUC) of the receiving operating
as post-translational protein modifications, regulation and xenobi- characteristic curve. For EGCG, network proximity offers good
otic metabolism (Fig. 1d). The enriched gene ontology categories discriminative power (AUC = 0.78, CI = 0.70–0.86) between dis-
indicate that polyphenols modulate common regulatory processes, eases with known and unknown therapeutic associations (Table 1).
but the low similarity in their protein targets, illustrated by the low We find that network proximity (dc) offers predictive power with
average JI, indicates that they target different processes within the an AUC > 0.7 for 31 polyphenols (Fig. 3b). The methodology
same process. recovers many associations well-documented in the literature,
We next asked whether the polyphenol targets cluster in specific such as the beneficial effects of umbelliferone on colorectal neo-
regions of the human interactome. We focused on polyphenols with plasms25,26. In Table 2, we summarize the top 10 polyphenols for
more than two targets (n = 46, Fig. 2) and measured the size and sig- which the network medicine framework offers the best predictive
nificance of the largest connected component (LCC) formed by the power of therapeutic effects, limiting the entries to those with pre-
targets of each polyphenol. We found that 25 of the 46 polyphenols dictive performance of AUC > 0.6 and where the precision of the
have a larger LCC than expected by chance (Z-score > 1.95) (Fig. 1e performance of the top predictions is greater than 0.6. Given the
and Fig. 2). In agreement with experimental evidence document- lack of data on true negative examples, we considered unknown
ing the effect of polyphenols on multiple pathways19, we find that associations as negative cases, observing the same trend when
ten polyphenols have their targets organized in multiple connected we used an alternative performance metric that does not require
components of size > 2. true negative labels (that is, AUC of the Precision–Recall curve)
These results indicate that the targets of polyphenols modulate (Supplementary Fig. 2).
specific well-localized neighbourhoods of the interactome (Fig. 2 Finally, we performed multiple robustness checks to exclude the
and Supplementary Fig. 1c). This prompted us to explore whether role of potential biases in the input data. To test whether the predic-
the interactome regions targeted by the polyphenols reside within tions are biased by the set of known associations retrieved from the
network neighbourhoods associated with specific diseases, thereby CTD, we randomly selected 100 papers from PubMed containing
seeking a network-based framework to unveil the molecular mecha- medical subject headings (MeSH) terms that tag EGCG to diseases.
nisms through which specific polyphenols modulate health. We manually curated the evidence for EGCG’s therapeutic effects
for the diseases discussed in the published papers, excluding reviews
Proximity between polyphenol targets and disease proteins and non-English language publications. The dataset was processed
reveals their therapeutic effects. Polyphenols can be viewed as to include implicit associations (Methods), resulting in a total of
drugs in that they bind to specific proteins, affecting their ability 113 diseases associated with EGCG, of which 58 overlap with the
to perform their normal functions. We therefore hypothesized that associations reported by the CTD (Fig. 3c). We observed that the
we can apply the network-based framework used to predict the predictive power of network proximity was unaffected by whether
efficacy of drugs in specific diseases16,17 to also predict the thera- we considered the annotations from the CTD, the manually curated
peutic effects of polyphenols. The closer the targets of a polyphe- list or the union of both (Fig. 3d). To test the role of potential biases
nol are to disease proteins, the more likely that the polyphenol in the interactome, we repeated our analysis using only high-quality
will affect the disease phenotype. We therefore calculated the net- polyphenol–protein interactions retrieved from ligand–protein
work proximity between polyphenol targets and proteins associ- three-dimensional resolved structures (Supplementary Fig. 1d)
ated with 299 diseases using the closest measure, dc, representing and a subset of the interactome derived from an unbiased
the average shortest path length between each polyphenol target high-throughput screening (Supplementary Fig. 1f). We found
and the nearest disease protein (Methods). Consider, for example, that the predictive power was largely unchanged, indicating that
(−)-epigallocatechin-3-O-gallate (EGCG), a polyphenol abundant the literature bias in the interactome does not affect our findings.
in green tea. Epidemiological studies have found a positive relation- Finally, we retested the predictive performance by considering not
ship between green tea consumption and reduced risk of T2D20,21, only the therapeutic polyphenol–disease associations, but also the
144 Nature Food | VOL 2 | March 2021 | 143–155 | www.nature.com/natfoodNaTure FOOd Articles
a
Polyphenol neighbourhood
Protein
Disease neighbourhoods
Polyphenol target
Disease proteins
LCC
b
387 121 251
c
50
Food Biofluids Number of polyphenols 40
ate
gall
759 polyphenols 30
-O- -3
atro atechin
641 in PubChem 20
res igalloc
l
10
tin
512 not in CTD
rce
ver
nol
129 in CTD
Ep
(no therapeutic associations)
Que
Phe
(–)-
(with therapeutic associations)
0
0 20 40 60 125 150 175 200 225
65 in STITCH Number of binding proteins
(experimental evidence
of protein interactions)
d Protein autophosphorylation 70
e
Peptidyl-tyrosine phosphorylation 128
Quercetin > 30
Peptidyl-tyrosine modification
Peptidyl-tyrosine autophosphorylation 60 64
–log10 (adjusted P value)
Flavonoid metabolic process 25
32
Hormone metabolic process Number of targets
50
Flavonoid glucuronidation
LCC size
16 20
Cellular glucuronidation
Flavonoid biosynthetic process 3-Caffeoylquinic acid
40 8
Uronic acid metabolic process Isoliquiritigenin 15
Glucoronate metabolic process
4
Regulation of hormone levels 30
Cellular hormone metabolic process 10
2
Positive regulation of kinase activity
Steroid metabolic process 20 RA
1Articles NaTure FOOd
MAP3K20
Quercetin LIMK2
PIK3CD
F2
PI4KA
ABCG2
PKN1
SRMS
KDM1A LIMK1
UGT1A1
ERN1 UGT1A8 UGT1A6
EPHB3
AKR1C4 PLAU EPHB6
EPHB4
IRAK3 FLT3
GCLM
PIK3C2B PIK3CG EPHA8
UGT1A7 UGT1A4
AKR1C2 ATP5F1A MAP3K7
AKR1C3
PTK6 TYK2
RIPK2 PIM1 ROR1
UGT1A9
ADORA2A
LCK EPHB1
IRAK1 JAK1 UGT1A3
AKR1D1 IRAK4
AXL
TEC MOS
EGFR
UGT1A10
AKR1C1 PLK1 RIPK1
AKR1B10 ATP5F1C PIK3CA JAK3
BTK ZAP70 SYK
IGF1R MAP3K21
FYN MAP3K10
AKR1B15 IRAK2 AURKB CDC42BPB
ITK
PIK3C2A INSRR
BLK LYN FGFR2
AKR1B1 ATP5F1B BMX PDGFRA PYGM
CALM1 PDGFRB
MET MAOA
PIK3CB UGT2B15 TEK
HSD3B1
INSR
TYRO3 PYGB
CDK1 EPHA3 TXK KIT ABL2
TRIB2 NUAK1 PIK3C3
PRKDC
CSF1R
AKT1 JAK2
MTOR FGR DRD4
CSK MATK MAOB HSD3B2
STK40 TRIB3
STK11 ABL1 TIE1
UGT2A3
PBK ATR KDR
FLT1
EPHA1 FRK
RIPK3
PTK2
EPHA2 FGFR1 SRC
TRIB1 EPHA4
GSK3A HCK PYGL CYP19A1
MELK ALOXE3
GPR35 MLKL
STK17B UXS1 YES1
ALOX12 GSK3B RET EPHB2 ALOX12B
CYP1A1
EPHA7
(–)-Epicatechin-3-O-gallate
NT5E FGFR4
UGT2B7
MAP3K11 PASK
SMG1
BCL2 ALOX5
CBR4 CAMK2B FLT4
NSDHL
STYK1 FGFR3
HSD17B8 MMP2
MAPK14
PIM2
CYP1A2 TMPRSS11D
HSD3B7
Myricetin
EPHA10 MMP9
(–)-Epigallocatechin-3-O-gallate Phenol
Butein
MTOR
MMP14
GZMB
AKR1C4 C1S
BACE1
MST1
TOP2A PRKDC
AKR1C2 KLK3 ST14
HSD3B2
MMP2 AKR1C3
HSD3B1 F12 ELANE
KLKB1 C2
DECR1 PLG
ATR
CBR4 PLAU HGF
DHRS7B PIK3CG AKR1C1 AKR1D1
HPGD CYP1A1 LPA
F10 PLAT MASP1
HPN F11 HGFAC
HSD17B8
APP ZFP36 AKR1B10 F2
NSDHL F7 MASP2
PIK3CD
ALOXE3 EGFR CELA1
CBR1
PROC
F9
COQ6 AKR1B15 PRSS46P
DHRSX CTRB2 CFB
CBR3
DECR2
3-Phenylpropionic acid PROZ NUF2 C1RL
PRKACA SNUPN
AKR1B1
ALOX12B GCLM
PTGS1 GOT1L1 XPO6
Quercetin 3-o-glucoside
PIN1 PRSS50 CTRB1 CFD
RAN HP
XPO1
DYRK1A NDC80
MAPT PTGS2 GOT1
BCL2
ALPG
Apigenin ESR2
ALPP
Chrysin XPO5
MAPK14
PYGL
GSK3B ALPI
3-Caffeoylquinic acid
ZFP36L1
FGFR1 Isoliquiritigenin Genistein
ESR1 AKR1C4
AKR1C4 ESR2 ESRRA
AR PYGM
KARS1
Caffeic acid
AKR1C2
Piceatannol JAK2
CDK6
PYGB
CYP1A1 AKR1C3
AKR1C3 CSNK2A1
AKR1C2 ESR1
ALK ALOX5
ACOT2
GRK5
AKR1D1 ABL1
MAP4K4 AKR1C1
ABCC1 EGFR
AKR1D1 AKR1C1
RET LTK
ACOT1 MAPK1
Cinnamic acid
TNK2 AKR1B10
RSPO4
AKR1B10 (–)-Epicatechin ESRRB
AKR1B15
Kaempeferol
UGT1A10
FLT1 RCOR1 ALPG
HCK UGT1A6
AKR1B15
AKR1B1
RSPO1 ALPP HSD17B1 UGT1A9
UGT2A3
UGT1A8
ALPI
AKR1B1
AHR
Resveratrol
ATP5F1B BTK UGT1A3
SYK LCK
RSPO3 KDM1A ALOX5 UGT1A1
SIRT3 LTA4H
UGT2B15
UGT1A7
TBK1 AKR1C4 UGT1A4 PPARG PIK3CA
ATP5F1C
ATP5F1A
Luteolin Ellagic acid INSR
SIRT1
ESR1
CA3
AKR1C2
LRRK2 CSNK1A1 AKR1C3
AKR1C4 ATP5F1B
IGF1R ATP5F1C
AKR1D1 KDR
STK3
FLT4 MET IQCB1
SIK2 AKR1B10 AKR1C2 AHRR
AKR1C3 SYK
SRC MYH14
AKR1C1 AKR1B1 SMAD3 PDGFRB
CAMK1D MYH8
AKR1B15
EGFR ATP5F1A
CBR4 AKT1 DUSP3 CYP2C9 MYO1C
PTGS1
MYH9
AKR1C1 MYH13
CSNK2A1 MYO6
AKR1D1 BRAF HSPA1B MYO5B
EPHB4 MYO1B
HSD17B8 CSNK2A1 MYH10
CSNK2A2 CA9
AKR1B10 GCLM CYP2C19
NUAK1 MYO5C MYO19 MYO5A
PTGS2
AKR1B15 MYO1E
CA2
AKR1B1
MYO18A
MYH11
MYO10
Fig. 2 | Protein–protein interactions of polyphenol targets. The 23 polyphenols whose targets form connected components in the interactome and
their respective subgraphs. For example, piceatannol targets form a unique connected component of 23 proteins, while quercetin targets form multiple
connected components, the largest having 140 proteins. Polyphenol targets that are not connected to any other target are not shown in the figure. Colours
distinguish connected components of different polyphenols.
marker/mechanism ones (another type of curated association avail- Network proximity predicts gene expression perturbation
able in the CTD) finding that the predictive power remains largely induced by polyphenols. To validate that network proximity reflects
unchanged (Supplementary Notes and Supplementary Fig. 3). the biological activity of polyphenols observed in experimental
146 Nature Food | VOL 2 | March 2021 | 143–155 | www.nature.com/natfoodNaTure FOOd Articles
a
MTNR1B
MMP14
EGCG
Found in MMP2
green tea
DECR2
DECR1
EGCG protein targets BACE1 MAPK8IP1
DHRSX
T2D disease proteins
APP
IRS1
MAPT
ZFP36 BCL11A
DYRK1A
SPRY2
DHRS7B
PIN1
COQ6
BLK
BCL2
MAPK14
HNF4A
PDX1 ZFP36L1
DUSP9
b c d
CTD Curation 1.0
15 0.8
Frequency
24 58 55 0.6
10
TPR
0.4
5
0.2
CTD AUC: 0.79, CI: 0.70–0.86
Curated AUC: 0.77, CI: 0.68–0.84
0 Combined AUC: 0.78, CI: 0.70–0.85
0.4 0.5 0.6 0.7 0.8 0.9 1.0 0 0.2 0.4 0.6 0.8 1.0
AUC FPR
Fig. 3 | Proximity between polyphenol targets and disease proteins is predictive of the therapeutic effects of the polyphenol. a, Interactome
neighbourhood showing the EGCG protein targets and their interactions with T2D-associated proteins. b, Distribution of AUC values of the predictions of
therapeutic effects for 65 polyphenols. c, Comparison of the EGCG–disease associations for the CTD and the in-house database derived from the manual
curation of the literature. d, Comparison of the prediction performance of the known EGCG–disease associations from the CTD, the in-house manually
curated database or the combined datasets. CI, confidence interval; TPR, true positive rate, FPR, false positive rate; dashed line, random expectation.
data, we retrieved expression perturbation signatures from the of disease genes among the most perturbed genes (Methods), find-
Connectivity Map database27 for the treatment of the breast cancer ing 13 diseases that have their genes significantly enriched among
MCF7 cell line with 21 polyphenols (Supplementary Table 1 and the genes most deregulated by genistein, of which four have known
Supplementary Fig. 4). We investigated the relationships between therapeutic associations. We find that these four diseases are sig-
the extent to which polyphenols perturb the expression of disease nificantly closer to the genistein targets than the nine diseases with
genes, the network proximity between the polyphenol targets and unknown therapeutic associations (Fig. 4c). We observed a similar
disease proteins and their known therapeutic effects (Fig. 4a). For trend for treatments with other polyphenols, whether we use the
example, we observe different perturbation profiles for gene pools same concentration (1 µM, Fig. 4c) or different ones (100 nM to
associated with different diseases: for treatment with genistein 10 µM, Supplementary Fig. 5). This result suggests that changes in
(1 µM, 6 hours) we observed 10 skin disease genes with perturbation gene expression caused by a polyphenol are indicative of its thera-
scores greater than 2, while we observed only one highly perturbed peutic effects, but only if the observed expression change is limited
cerebrovascular disorder gene (Fig. 4b). Indeed, network proxim- to proteins proximal to the polyphenol targets (Fig. 4a).
ity indicates that skin disease is closer to the genistein targets than Consequently, network proximity should also be predictive
cerebrovascular disorder, suggesting a relationship between net- of the overall gene expression perturbation caused by a polyphe-
work proximity, gene expression perturbation and the therapeutic nol on the genes of a given disease. To test this hypothesis, in each
effects of the polyphenol (Fig. 4a). To test this hypothesis, we com- experimental combination defined by the polyphenol type and its
puted an enrichment score that measures the over-representation concentration, we evaluated the maximum perturbation among
Nature Food | VOL 2 | March 2021 | 143–155 | www.nature.com/natfood 147Articles NaTure FOOd
result in therapeutic effects, suggesting that future studies could
Table 1 | Top 20 predicted therapeutic associations between integrate gene expression (whenever available) with network
EGCG and human diseases proximity as they aim to more accurately prioritize polyphenol–
Disease Distance Significance Known
disease associations.
dc Zdc therapeutic
effect Experimental evidence confirms that RA modulates platelet
(references) function. To demonstrate how the network-based framework can
facilitate the mechanistic interpretation of the therapeutic effects
Nervous system diseases 1.13 −1.72 64,65
of selected polyphenols, we next focus on VD. Of 65 polyphe-
Nutritional and metabolic 1.25 −1.45 23 nols evaluated in this study, we found 27 to have associations to
diseases VD, as their targets were within the VD network neighbourhood
Metabolic diseases 1.25 −1.41 23 (Supplementary Table 3). We therefore inspected the targets of 15 of
Cardiovascular diseases 1.27 −2.67 66–71 the 27 polyphenols with 10 or fewer targets. The network analysis
identified direct links between biological processes related to vascu-
Immune system diseases 1.29 −1.31 72
lar health and the targets of three polyphenols: gallic acid, RA and
Vascular diseases 1.33 −3.47 66,67,70 1,4-naphthoquinone (Supplementary Fig. 12 and Supplementary
Digestive system diseases 1.33 −1.57 73–77 Notes). The network neighbourhood containing the targets of these
Neurodegenerative diseases 1.37 −1.71 78
polyphenols suggests that gallic acid activity involves thrombus dis-
solution processes, RA acts on platelet activation and antioxidant
Central nervous system 1.41 −0.54 78 pathways through FYN and its neighbours and 1,4-naphthoquinone
diseases
acts on signalling pathways of vascular cells through MAP2K1
Autoimmune diseases 1.41 −1.30 72 activity (Supplementary Fig. 12 and Supplementary Notes).
Gastrointestinal diseases 1.43 −1.02 79 To validate the developed framework, we set out to obtain direct
Brain diseases 1.43 −0.89 NA
experimental evidence of the predicted mechanistic role of RA in
VD. The RA targets are in close proximity to proteins related to
Intestinal diseases 1.49 −1.08 79 platelet function, forming the RA/VD-platelet module: a connected
Inflammatory bowel diseases 1.54 −2.10 NA component formed by the RA target FYN and the VD proteins asso-
Bone diseases 1.54 −1.18 NA ciated with platelet function PDE4D, CD36 and APP (Fig. 6a). We
therefore asked whether RA influenced platelet activation in vitro.
Gastroenteritis 1.54 −1.92 NA
As platelets can be stimulated through different activation pathways,
Demyelinating diseases 1.54 −1.78 NA RA effects can, in principle, occur in any of them. To test these dif-
Glucose metabolism 1.54 −1.58 23 ferent possibilities, we pretreated platelets with RA and then either
disorders (1) activated glycoprotein VI by collagen or collagen-related pep-
Heart diseases 1.56 −1.20 68,69,71 tide (CRP/CRPXL); (2) activated protease-activated receptors-1,4
by thrombin receptor activator peptide-6 (TRAP-6); (3) activated
Diabetes mellitus 1.56 −1.66 23
prostanoid thromboxane receptor by the thromboxane A2 analogue
Diseases were ordered according to the network distance (dc) of their proteins to EGCG targets, (U46619) or (4) activated P2Y1/12 receptor by adenosine diphos-
and diseases with relative distance Z > −0.5 were removed. References reported in the CTD
dc
for curated therapeutic associations are shown. NA indicates diseases with no documented
phate (ADP)28. When we compared the network distance between
therapeutic association in CTD. each stimulant receptor and the RA/VD-platelet module (Fig. 6a), we
observed that the receptors for CRP/CRPXL, TRAP-6 and U46619
are closer than would be expected for a random distribution, while
genes for each disease. We then compared the magnitude of the the receptor for ADP is more distant (Fig. 6b). We expected that
observed perturbation between diseases that were proximal (dc platelets would be most affected by RA when treated with stimu-
below the 25th percentile, Zdc < −0.5) or distal (dc above the 75th lants whose receptors are most proximal to the RA/VD-platelet
percentile, Zdc > −0.5) to the polyphenol targets. Figure 5a,b and module, that is, CRP/CRPXL, TRAP-6 and U46619, and as a con-
Supplementary Fig. 6 show the results for the genistein treatment trol, we expect no effect for the distant ADP receptor. The experi-
(1 µM, 6 hours), which indicate that diseases proximal to the poly- ments confirm this prediction: RA inhibits collagen-mediated
phenol targets show higher maximum perturbation values than dis- platelet aggregation (Fig. 6c) and impairs dense granule secretion
tal diseases. The same trend is observed for other polyphenols when induced by CRPXL, TRAP-6 and U46619 (Supplementary Fig. 13).
we use different dc and Zdc thresholds for defining proximal and RA-treated platelets also displayed dampened α-granule secretion
distant diseases (Fig. 5b and Supplementary Figs. 6–9), confirm- (Fig. 6d) and integrin αIIbβ3 activation (Supplementary Fig. 13) in
ing that the impact of a polyphenol on cellular signalling pathways response to U46619. As expected, RA did not affect platelet func-
is localized in the network space, being greater in the vicinity of tion when we used an agonist whose receptor is distant from the
the polyphenol targets than in neighbourhoods remote from these RA/VD-platelet module (that is, ADP). These findings suggest that
targets. We also considered gene expression perturbations in the RA impairs basic hallmarks of platelet activation via strong network
network vicinity of the polyphenol targets, regardless of whether effects, supporting our hypothesis that the proximity between RA
the proteins were disease proteins, and observed higher perturba- targets and the neighbourhood associated with platelet function
tion scores for proximal proteins in 12 out 21 polyphenols tested at (Fig. 6a) could in part explain RA’s impact on VD.
10 µM (Supplementary Fig. 10). Finally, we found that the enrich- We next sought to clarify the molecular mechanisms involved in
ment score of perturbed genes among disease genes was not as pre- the impact of RA on platelets. Given that platelet activation is coor-
dictive of the polyphenol therapeutic effects as network proximity dinated by several kinases, we hypothesized that RA inhibits platelet
(Supplementary Fig. 11). function by blocking agonist-induced protein tyrosine phosphor-
Altogether, these results indicate that network proximity offers ylation. We observed that RA-treated platelets demonstrated a
a mechanistic interpretation for the gene expression perturbations dose-dependent reduction in total tyrosine phosphorylation in
induced by polyphenols on disease genes. They also show that net- response to CRPXL, TRAP-6 and U46619 (Fig. 6e). Given that RA
work proximity can indicate when gene expression perturbations caused a substantial decrease in phosphorylation of proteins with
148 Nature Food | VOL 2 | March 2021 | 143–155 | www.nature.com/natfoodNaTure FOOd Articles
Table 2 | Top-ranked polyphenols
Polyphenol AUC AUC CIa Precisionb Concentration in bloodc (μM) No. of mapped targets LCC size
Coumarin 0.93 [0.86–0.98] 0.6 7 1
Piceatannol 0.86 [0.77–0.94] 0.6 39 23
Genistein 0.82 [0.75–0.89] 0.7 [0.006–0.525] 18 6
Ellagic acid 0.79 [0.63–0.92] 0.6 42 19
EGCG 0.78 [0.70–0.86] 0.8 51 17
Isoliquiritigenin 0.75 [0.77–0.94] 0.6 10 8
Resveratrol 0.75 [0.66–0.82] 1 63 25
Pterostilbene 0.73 [0.61–0.84] 0.6 5 2
Quercetin 0.73 [0.64–0.81] 1 [0.022–0.080] 216 140
(−)-Epicatechin 0.65 [0.49–0.80] 0.8 0.625 11 3
Polyphenols for which prediction of therapeutic effects by network proximity to diseases was most successful. The table shows polyphenols with AUC > 0.6 and precision > 0.6. aCIs calculated with 2,000
bootstraps with replacement and sample size of 50% of the diseases (150 / 299). bPrecision was calculated based on the top 10 polyphenols after their ranking based on the distance (dc) of their targets to
the disease proteins and considering only predictions with Z-score < −0.5. cConcentrations of polyphenols in blood were retrieved from the Human Metabolome Database (HMDB).
atomic mass between 50 and 60 kDa (Fig. 6e), we hypothesized that The lack of these data does not invalidate the findings presented
RA may reduce phosphorylation of FYN (59 kDa) or other similarly here, since previous studies report the presence of unmetabolized
sized members of the same protein family (that is, SFKs). To test polyphenols in blood31–33 and it has been hypothesized that, in
this, we measured the level of phosphorylation within the activa- some instances, deconjugation of liver metabolites occurs in spe-
tion domain (amino acid 416) of SFKs, finding that RA reduced cific tissues or cells34–36. Therefore, the lack of data concerning spe-
collagen induced phosphorylation of FYN as well as basal tyrosine cific polyphenols and the fact that other mechanisms exist through
phosphorylation of SFKs (Fig. 6f). This indicates that RA perturbs which they can affect health (for example, antioxidant activity and
the phospho-signalling networks that regulate platelet response to microbiota regulation) explain why this methodology might still
extracellular stimuli. miss a few known relationships between polyphenols and diseases.
Taken together, these findings support our prediction that RA Second, considering that several experimental studies of polyphe-
modulates platelet activation and function. They also support the nol bioefficacy have been observed in in vitro and in vivo models,
observation that its mechanism of action involves reduction of the proposed framework might help us interpret literature evidence,
phosphorylation at the activation domain of the protein tyrosine possibly even allowing us to exclude chemical candidates when con-
kinase FYN (Fig. 6a) and the inhibition of general tyrosine phos- sidering the health benefits provided by a given food in epidemio-
phorylation. Finally, while polyphenols are usually associated with logical association studies.
their antioxidant function, here we illustrate another mechanistic Our assumption that network proximity recovers thera-
pathway through which they could benefit health. peutic associations is based on its predictive performance on a
ground-truth dataset for observed therapeutic effects and also relies
Discussion on previous observations about the effect of drugs on diseases16,17,37.
Here, we propose a network-based framework to predict the thera- While the proposed methodology offers a powerful prioritization
peutic effects of dietary polyphenols in human diseases. We find tool to guide future research, the real effect of polyphenols on dis-
that polyphenol protein targets cluster in specific functional neigh- eases might still be negative, given other unmet factors such as dos-
bourhoods of the interactome, and we show that the network prox- age, comorbidities and drug interactions, which can only be ruled
imity between polyphenol targets and disease proteins is predictive out by preclinical and clinical studies. Gene expression perturba-
of the therapeutic effects of polyphenols. We demonstrate that dis- tion profiles, such as the ones provided by the Connectivity Map,
eases whose proteins are proximal to polyphenol targets tend to have can also be integrated with network proximity to further highlight
significant changes in gene expression in cell lines treated with the potential beneficial or harmful effects of chemical compounds38,39.
respective polyphenol, while such changes are absent for diseases The low bioavailability of some polyphenols in food might still
whose proteins are distal to polyphenol targets. Finally, we find that present challenges when considering the therapeutic utility of these
the network neighbourhood around the RA targets and VD pro- molecules. However, 48 of the 65 polyphenols we explored here are
teins are related to platelet function. We validate this mechanistic predicted to have high gastrointestinal absorption (Supplementary
prediction by showing that RA modulates platelet function through Table 2) and different methodologies are available to increase bio-
inhibition of protein tyrosine phosphorylation. These observations availability of natural compounds40,41. Additionally, in the same
suggest a role of RA on prevention of VD by inhibiting platelet acti- way that the polyphenol phlorizin led to the discovery of new
vation and aggregation. strategies for disease treatment resulting in the development of
The observed results also suggest multiple avenues through new compounds with higher efficacy42, we believe that the present
which our ability to understand the role of polyphenols could be methodology can help us identify polyphenol-based candidates for
improved. First, some of the known health benefits of polyphenols drug development.
might be caused not only by the native molecules but also by their The methodology introduced here offers a foundation for the
metabolic byproducts29,30. However, we lack data concerning colonic mechanistic interpretation of alternative pathways through which
degradation, liver metabolism, bioavailability and interaction with polyphenols can affect health, such as the combined effect of differ-
proteins of specific polyphenols or their metabolic byproducts. ent polyphenols37,43 and their interactions with drugs44. To address
Future experimental data on protein interactions with polyphenol such synergistic effects, we need ground-truth data on these aspects.
byproducts and conjugates could be incorporated in the proposed The developed methodology can be applied to other food-related
framework, further improving the accuracy of our predictions. chemicals, providing a framework by which to understand their
Nature Food | VOL 2 | March 2021 | 143–155 | www.nature.com/natfood 149Articles NaTure FOOd
a Disease distant and with no enrichment
of perturbed genes
Effect
Polyphenol targets
unknown
Protein
Gene expression perturbed
by polyphenol treatment
Disease distant
and with enrichment
of perturbed genes
Therapeutic
b Disease proximal and
with enrichment of perturbed genes
RAD51D
Skin diseases (therapeutic)
Disease genes (proximal)
GZMB
Cerebrovascular disorders (effect unknown)
Disease genes (distant)
Genistein targets
NOP10
Perturbed genes
RAD51C
ERCC3
ESRRA
ESR2 CCL1
CAST ABL1
EFNA5
ESR1
APP
NR1H2 PDE4D EPHA4
EGFR
KTR13 ESRRB ACE
TTR
HTR2A
c Genistein (1 µM) Quercetin (1 µM) Resveratrol (1 µM) Myricetin (1 µM)
2.0
2.0 –5 2.0 P = 7.45 × 10–3
P = 1.58 × 10 2.0
Closest
Closest
Closest
Closest
1.6
1.6 1.6 1.6
1.2 P = 4.41 × 10–2 P = 8.50 × 10–6 1.2
1.2 1.2
Effect Therapeutic Effect Therapeutic Effect Therapeutic Effect Therapeutic
unknown (n = 25) (n = 14) unknown (n = 7) (n = 10) unknown (n = 37) (n = 59) unknown (n = 84) (n = 13)
Fig. 4 | Relationships among gene expression perturbation, network proximity and the therapeutic effects of polyphenols on diseases. a, Schematic
representation of the relationships between the extent to which a polyphenol perturbs disease genes expression, its proximity to the disease genes and
its therapeutic effects. b, Interactome neighbourhood showing the modules of skin diseases, genistein and cerebrovascular disorders. The skin diseases
module has 10 proteins with high perturbation scores (>2) in the treatment of the MCF7 cell line with 1 µM of genistein. Genes associated with skin
disease are significantly enriched among the most differentially expressed genes, and the maximum perturbation score among disease genes is higher
in skin disease than cerebrovascular disorders. c, Among the diseases in which genes are enriched with highly perturbed genes, those with therapeutic
associations show smaller network distances to the polyphenol targets than those without. The same trend is observed in treatments of the polyphenols
quercetin, resveratrol and myricetin. Boxplots show the median (horizontal line), 25th and 75th percentiles (lower and upper boundaries, respectively).
Whiskers extend to data points that lie within 1.5 interquartile ranges of the 25th and 75th quartiles; and observations that fall outside this range are
displayed independently.
150 Nature Food | VOL 2 | March 2021 | 143–155 | www.nature.com/natfoodNaTure FOOd Articles
a
Therapeutic
Known
Unknown 7
3
Maximum perturbation score
4
2
5
Disease labels Genistein 1
0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
1: Nervous system diseases
Network proximity (dc)
2: Genetic diseases, inborn
3: Metabolism, inborn errors
4: Congenital, hereditary, and neonatal diseases 6
10
5: Neoplasms
6: Colonic diseases
7: Macular degeneration
8
8: Spondylitis, ankylosing
9: Dyssomnias 9
10: Osteoporosis
b Genistein (1 µM) Resveratrol (1 µM)
1.0 Proximal (n = 48)
1.0 Proximal (n = 37)
Distant (n = 48) Distant (n = 54)
0.8 0.8
0.6 0.6
CDF
CDF
0.4 0.4
0.2 0.2
P = 4.25 × 10–9 P = 1.31 × 10–12
0 0
0 4 8 0 2 4
Maximum perturbation score Maximum perturbation score
Quercetin (1 µM) Myricetin (1 µM)
1.0 1.0 Proximal (n = 51)
Distant (n = 57)
0.8 0.8
0.6 0.6
CDF
CDF
0.4 0.4
–16
P = 8.88 × 10
0.2 0.2
Proximal (n = 61)
Distant (n = 67) P = 1.31 × 10–12
0 0
0 3 6 0.5 2 3.5
Maximum perturbation score Maximum perturbation score
Fig. 5 | Diseases proximal to polyphenol targets have higher gene expression perturbation profiles. a, Proximal and distal diseases in relation to genistein
targets. Each node represents a disease and the node size is proportional to the perturbation score after treatment with genistein (1 µM, 6 hours). Distance
from the origin represents the network proximity (dc) to genistein targets. Purple nodes represent diseases for which the therapeutic association was
previously known. b, Cumulative distribution function (CDF) of the maximum perturbation scores of genes from diseases that are distal or proximal to
polyphenol targets for the polyphenols genistein, quercetin, resveratrol and myricetin (1 µM, 6 hours). Statistical significance was evaluated with the
Kolmogorov–Smirnov test.
Nature Food | VOL 2 | March 2021 | 143–155 | www.nature.com/natfood 151Articles NaTure FOOd
a ADP b
FCGR2A
receptor
CRP/CRPXL GP6
receptor P2RY12
Average shortest path length
GNAI2
Random
GNB1 2.8
TBXA2R GNAQ
FYN RAF1
F2R
2.4
CD36
PRKCA
JAK2
EPHA4
TRAP-6
IRS1 APP GNA12 receptor 2.0
PDE4D
VD proteins
VHL
U46619
GNA15
RA targets IER3 receptor 1.6
GP6 TBXA2R F2R ADP
Neighbouring proteins MCL1
RA/VD-platelet module
AKR1B1 Receptors of platelet stimulants
TNKS
c d 100 100
100 100
75 75
75 75
50 50 50 50
CD62p positive platelets (%)
Maximum aggregation (%)
25 25 25 25
CRP TRAP-6 P = 0.53 CRPXL TRAP-6 P = 0.32
P = 0.005 P = 0.78
0 0 0 0
0 10 100 1,000 0 10 100 1,000 0 10 100 1,000 0 10 100 1,000
100 100
100 100
75 75
75 75 P = 0.003 P = 0.11
50 50 50 50
25 25 25 25
U46619 P = 0.18 ADP P = 0.64 U46619 ADP
0 0 0 0
0 10 100 1,000 0 10 100 1,000 0 10 100 1,000 0 10 100 1,000
RA concentration (µM) RA concentration (µM)
e f
kDa 0 µM RA 100 µM RA 500 µM RA IP: FYN
SFK (p-Tyr416) SFK (p-Tyr416)
Rest CRPXL T6 U4 CRPXL T6 U4 CRPXL T6 U4
2.5 2.0
180 P = 0.02
P = 0.24
82 2.0 P = 0.05
Relative band pixel density
Relative band pixel density
64 1.5
1.5
48 P = 0.21
4G10
P-Tyr
37 1.0
0.5
26
0.5
19
0 0
RA – + – + + +
– –
(500 µM)
Collagen – – + + – – + +
(1 µg ml–1)
Fig. 6 | RA modulates platelet function. a, Interactome neighbourhood showing RA targets and the RA/VD-platelet module (that is, the connected
component formed by the RA target FYN and the VD proteins associated with platelet function (PDE4D, CD36 and APP)) and the receptor for platelet
agonists used in our experiments (collagen/CRPXL, TRAP-6, U46619 and ADP). b, Average shortest path length from each platelet agonist receptor and
the RA/VD-platelet module formed by the proteins FYN, PDE4D, CD36 and APP. Error bars represent standard deviation of that same measure over
1,000 iterations of random selection of nodes in a degree-preserving fashion. c–f, Platelet-rich plasma (PRP) or washed platelets were pretreated with RA
for 1 hour before stimulation with either collagen (1 μg ml−1), collagen-related peptide (CRPXL, 1 μg ml−1), thrombin receptor activator peptide-6 (TRAP-6,
20 μM), U46619 (1 μM) or ADP (10 μM). Platelets were assessed for either aggregation (c) or α-granule secretion (d). Platelet lysates were also probed for
either non-specific tyrosine phosphorylation (p-Tyr) of the whole cell lysate (e) or site-specific phosphorylation of Src family kinases (SFKs) and FYN at
residue 416 (f). n = 3–6 separate blood donations. In c and d, points represent the mean and error bars represent ± s.e.m. In f, points represent values from
biological replicates and error bars represent s.e.m. P values were determined by the Kruskal–Wallis test in c and d and by unpaired t-tests in f.
152 Nature Food | VOL 2 | March 2021 | 143–155 | www.nature.com/natfoodNaTure FOOd Articles
d − μdc (S,T)
health effects. Future research may help us also account for the Z dc = (2)
σ dc (S,T)
way food-related chemicals affect endogenous metabolic reactions,
impacting not only signalling pathways but also catabolic and ana- We performed a degree-preserving random selection, but due to the scale-free
bolic processes. Finally, the methodology provides a framework nature of the human interactome, we avoid repeatedly choosing the same
to interpret and find causal support for associations identified in (high-degree) nodes by using a binning approach in which nodes within a certain
observational studies. Taken together, the proposed network-based degree interval were grouped together such that there were at least 100 nodes in the
bin. Supplementary Data 2 reports the proximity scores dc and Zdc for all pairs of
framework has the potential to systematically reveal the mechanism
diseases and polyphenols.
of action underlying the health benefits of polyphenols, offering a
logical, rational strategy for mechanism-based drug development of Area under receiving operating characteristic curve analysis. For each
food-based compounds. polyphenol, we used the AUC to evaluate how well the network proximity
distinguishes diseases with known therapeutic associations from all the others
in the set of 299 diseases. The set of known associations (therapeutic) retrieved
Methods from the CTD were used as positive instances and all unknown associations
Building the interactome. The human interactome was assembled from 16
were defined as negative instances, and the area under the receiving operating
databases containing six different types of protein–protein interactions (PPIs):
characteristic curve was computed using the implementation in the Scikit-learn
(1) binary PPIs tested by high-throughput yeast two-hybrid (Y2H) experiments45;
Python package. Furthermore, we calculated 95% CIs using the bootstrap
(2) kinase–substrate interactions from literature-derived low-throughput and
technique with 2,000 resamplings with sample sizes of 150 each. Considering that
high-throughput experiments from KinomeNetworkX46, Human Protein Resource
the AUC provides an overall performance, we also searched for a metric to evaluate
Database (HPRD)47 and PhosphositePlus48; (3) carefully literature-curated PPIs
the top-ranking predictions. For this analysis, we calculated the precision of the top
identified by affinity purification followed by mass spectrometry (AP-MS) and
10 predictions, considering only the polyphenol–disease associations with relative
from literature-derived low-throughput experiments from InWeb49, BioGRID50,
distance Zdc < −0.5 (ref. 16).
PINA51, HPRD52, MINT53, IntAct53 and InnateDB54; (4) high-quality PPIs from
three-dimensional protein structures reported in Instruct55, Interactome3D56
and INSIDER57; (5) signalling networks from literature-derived low-throughput Analysis of network proximity and gene expression deregulation. We retrieved
experiments as annotated in SignaLink2.0 (ref. 58) and (6) protein complexes perturbation signatures from the Connectivity Map database (https://clue.io/) for
from BioPlex2.0 (ref. 59). The genes were mapped to their Entrez ID based on the the MCF7 cell line after treatment with 21 polyphenols. These signatures reflect
National Center for Biotechnology Information (NCBI) database as well as their the perturbation of the gene expression profile caused by the treatment with that
official gene symbols. The resulting interactome includes 351,444 PPIs connecting particular polyphenol relative to a reference population, which comprises all
17,706 unique proteins (Supplementary Data 1). The LCC has 351,393 PPIs and other treatments in the same experimental plate27. For polyphenols having more
17,651 proteins. than one experimental instance (such as time of exposure, cell line and dose),
we selected the one with highest distil_cc_q75 value (75th quantile of pairwise
Polyphenols, polyphenol targets and disease proteins. We retrieved 759 Spearman correlations in landmark genes, https://clue.io/connectopedia/glossary).
polyphenols from the PhenolExplorer database4. The database lists polyphenols We performed gene set enrichment analysis61 to evaluate the enrichment of
with food composition data or that have been profiled in biofluids after disease genes among the top deregulated genes in the perturbation profiles. This
interventions with polyphenol-rich diets. For our analysis, we only considered analysis offers enrichment scores that have small values when genes are randomly
polyphenols that (1) could be mapped in PubChem IDs, (2) were listed in the distributed among the ordered list of expression values and high values when they
Comparative Toxicogenomics (CTD) database24 as having therapeutic effects on are concentrated at the top or bottom of the list. The enrichment score significance
human diseases and (3) had protein-binding information present in the STITCH is calculated by creating 1,000 random selections of gene sets with the same size as
database60 with experimental evidence (Fig. 1a). After these steps, we considered the original set and calculating an empirical P value by considering the proportion
a final list of 65 polyphenols, for which 598 protein targets were retrieved of random sets resulting in enrichment scores smaller than the original case.
from STITCH (Supplementary Table 1). We considered 3,173 disease proteins The P values were adjusted for multiple testing using the Benjamini–Hochberg
corresponding to 299 diseases retrieved from Menche et al.15. Gene ontology method. The network proximity dc of disease proteins and polyphenol targets for
enrichment analysis of protein targets was performed using the Bioconductor diseases with significant enrichment scores were compared according to their
package clusterProfiler with a significance threshold of P < 0.05 and Benjamini– therapeutic and unknown-therapeutic associations using the Student’s t-test.
Hochberg multiple testing correction with Q < 0.05. The relevant code for calculating the network proximity, AUCs and enrichment
scores can be found at https://github.com/italodovalle/polyphenols.
Polyphenol–disease associations. We retrieved the polyphenol–disease
associations from the Comparative Toxicogenomics Database (CTD). We Platelet isolation. Human blood collection was performed as previously
considered only manually curated associations labelled as therapeutic. By described in accordance with the Declaration of Helsinki and ethics regulations
considering the hierarchical structure of diseases along the MeSH tree, we with Institutional Review Board approval from Brigham and Women’s Hospital
expanded explicit polyphenol–disease associations to also include implicit (P001526). Healthy volunteers did not ingest known platelet inhibitors for at
associations. This procedure was performed by propagating associations in least 10 days prior. Citrated whole blood underwent centrifugation with a slow
the lower branches of the MeSH tree to consider diseases in the higher levels brake (177 × g, 20 minutes), and the PRP fraction was acquired for subsequent
of the same tree branch. For example, a polyphenol associated with heart experiments. For washed platelets, PRP was incubated with 1 μM prostaglandin
diseases would also be associated with the more general category of cardiovascular E1 (Sigma, P5515) and immediately underwent centrifugation with a slow brake
diseases. By performing this expansion, we obtained a final list of 1,525 (1,000 × g, 5 minutes). Platelet-poor plasma was aspirated, and pellets were
known associations between the 65 polyphenols and the 299 diseases resuspended in platelet resuspension buffer (PRB; 10 mM HEPES, 140 mM NaCl,
considered in this study. 3 mM KCl, 0.5 mM MgCl2, 5 mM NaHCO3, 10 mM glucose, pH 7.4).
Network proximity between polyphenol targets and disease proteins. Platelet aggregometry. Platelet aggregation was measured by turbidimetric
The proximity between a disease and a polyphenol was evaluated using a aggregometry as previously described62. Briefly, PRP was pretreated with RA
distance metric that takes into account the shortest path lengths between for 1 hour before adding 250 μl to siliconized glass cuvettes containing magnetic
polyphenol targets and disease proteins16. Given S, the set of disease proteins, stir bars. Samples were placed in Chrono-Log Model 700 Aggregometers before
T, the set of polyphenol targets and d(s,t), the shortest path length between the addition of various platelet agonists. Platelet aggregation was monitored for
nodes s and t in the network, we define: 6 minutes at 37 °C with a stir speed of 1,000 r.p.m. and the maximum extend of
aggregation recorded using AGGRO/LINK8 software. In some cases, dense granule
1 ∑ release was simultaneously recorded by supplementing samples with Chrono-Lume
dc (S, T) = min d (s, t) (1)
∥ T∥ t∈T s∈S (Chrono-Log, 395) according to the manufacturer’s instructions.
We also calculated a relative distance metric ( Z ) that compares the absolute Platelet α-granule secretion and integrin αIIbβ3 activation. Changes in
d
distance dc(S,T) between a disease and a polyphenolc with a reference distribution platelet surface expression of P-selectin (CD62P) or binding of Alexa Fluor
describing the random expectation. The reference distribution corresponds to the 488-conjugated fibrinogen were used to assess α-granule secretion and integrin
expected distances between two randomly selected groups of proteins matching αIIbβ3 activation, respectively. First, PRP was preincubated with RA for 1 hour,
the size and degrees of the original disease proteins and polyphenol targets in followed by stimulation with various platelet agonists under static conditions
the network. It was generated by calculating the proximity between these two at 37 °C for 20 minutes. Samples were then incubated with APC-conjugated
randomly selected groups across 1,000 iterations. The mean μd(S,T) and standard anti-human CD62P antibodies (BioLegend, 304910) and 100 μg ml−1 Alexa Fluor
deviation σd(S,T) of the reference distribution were used to convert the absolute 488-Fibrinogen (Thermo Scientific, F13191) for 20 minutes before fixation in
distance dc into the relative distance Zdc, defined as: 2% (v/v) paraformaldehyde (Thermo Scientific, AAJ19945K2). For each sample,
Nature Food | VOL 2 | March 2021 | 143–155 | www.nature.com/natfood 153Articles NaTure FOOd
50,000 platelets were processed using a Cytek Aurora spectral flow cytometer. 13. Bhatt, J. K., Thomas, S. & Nanjan, M. J. Resveratrol supplementation
Percent positive cells were determined by gating on fluorescence intensity improves glycemic control in type 2 diabetes mellitus. Nutr. Res. 32,
compared to unstimulated samples. 537–541 (2012).
14. Sharma, A. et al. A disease module in the interactome explains disease
Platelet cytotoxicity. Cytotoxicity was tested by measuring lactate dehydrogenase heterogeneity, drug response and captures novel pathways and genes in
(LDH) release by permeabilized platelets into the supernatant63. Briefly, washed asthma. Hum. Mol. Genet. 24, 3005–3020 (2014).
platelets were treated with various concentrations of RA for 1 hour, before isolating 15. Menche, J. et al. Disease networks. Uncovering disease–disease relationships
supernatants via centrifugation (15,000 × g, 5 min). A Pierce LDH Activity Kit through the incomplete interactome. Science 347, 1257601 (2015).
(Thermo Scientific, 88953) was then used to assess supernatant levels of LDH. 16. Guney, E., Menche, J., Vidal, M. & Barabási, A.-L. Network-based in silico
drug efficacy screening. Nat. Commun. 7, 10331 (2016).
Immunoprecipitation and western blot. Washed platelets were pretreated with 17. Cheng, F. et al. Network-based approach to prediction and population-based
RA for 1 hour, followed by a 15 minute treatment with Eptifibatide (50 μM). validation of in silico drug repurposing. Nat. Commun. 9, 2691 (2018).
Platelets were then stimulated with various agonists for 5 minutes under stirring 18. Kovács, I. A. et al. Network-based prediction of protein interactions.
conditions (1,000 r.p.m., 37 °C). Platelets were lysed on ice with RIPA Lysis Buffer Nat. Commun. 10, 1240 (2019).
System (Santa Cruz, sc-24948) and supernatants clarified via centrifugation 19. Sarkar, F. H., Li, Y., Wang, Z. & Kong, D. Cellular signaling perturbation by
(15,000 × g, 10 min, 4 °C). For immunoprecipitation of FYN, lysates were first natural products. Cell. Signal. 21, 1541–1547 (2009).
precleared of IgG by incubating with Protein A agarose beads (Cell Signaling 20. Iso, H. et al. The relationship between green tea and total caffeine intake and
Technologies, 9863S) for 30 minutes at 4 °C, before isolation of the supernatant risk for self-reported type 2 diabetes among Japanese adults. Ann. Intern.
via centrifugation (15,000 × g, 10 min, 4 °C). Supernatants were incubated with Med. 144, 554–562 (2006).
anti-FYN antibodies (Abcam, 2A10) overnight at 4 °C before incubation with 21. Song, Y., Manson, J. E., Buring, J. E., Sesso, H. D. & Liu, S. Associations of
Protein A beads for 1 hour. Beads were then washed five times with NP-40 lysis dietary flavonoids with risk of type 2 diabetes, and markers of insulin
buffer (144 mM Tris, 518 mM NaCl, 6 mM EDTA, 12 mM Na2VO3, 33.3% (v/v) resistance and systemic inflammation in women: a prospective study and
NP-40, Halt protease inhibitor cocktail (Thermo, 78429)). cross-sectional analysis. J. Am. Coll. Nutr. 24, 376–384 (2005).
For western blot analysis, total cell lysates or immunoprecipitated FYN were 22. Keske, M. A. et al. Vascular and metabolic actions of the green tea
reduced with Laemmli Sample Buffer (Bio-Rad, 1610737) and proteins separated polyphenol epigallocatechin gallate. Curr. Med. Chem. 22, 59–69 (2015).
by molecular weight in PROTEAN TGX precast gels (Bio-Rad, 4561084). Proteins 23. Wolfram, S. et al. Epigallocatechin gallate supplementation alleviates diabetes
were transferred to PVDF membranes (Bio-Rad, 1620174) and probed with either in rodents. J. Nutr. 136, 2512–2518 (2006).
4G10 (Milipore, 05-321), a primary antibody clone that recognizes phosphorylated 24. Davis, A. P. et al. The comparative toxicogenomics database: update 2019.
tyrosine residues, or primary antibodies that probe for the site-specific Nucleic Acids Res. 47, D948–D954 (2019).
phosphorylation of SFKs (p-Tyr416) within their activation loop. Membranes were 25. Muthu, R., Selvaraj, N. & Vaiyapuri, M. Anti-inflammatory and proapoptotic
incubated with horseradish peroxidase-conjugated secondary antibodies (Cell effects of umbelliferone in colon carcinogenesis. Hum. Exp. Toxicol. 35,
Signaling Technologies, 7074S) to catalyse an electrochemiluminescent reaction 1041–1054 (2016).
(Thermo Scientific, PI32109). Membranes were visualized using a Bio-Rad 26. Muthu, R. & Vaiyapuri, M. Synergistic and individual effects of umbelliferone
ChemiDoc Imaging System, and densitometric analysis of protein lanes was with 5-fluorouracil on tumor markers and antioxidant status of rat treated
conducted using ImageJ (NIH, Version 1.52a). with 1,2-dimethylhydrazine. Biomed. Aging Pathol. 3, 219–227 (2013).
27. Subramanian, A. et al. A next generation connectivity map: L1000 platform
Reporting Summary. Further information on research design is available in the and the first 1,000,000 profiles. Cell 171, 1437–1452.e17 (2017).
Nature Research Reporting Summary linked to this article. 28. Grover, S. P., Bergmeier, W. & Mackman, N. Platelet signaling pathways and
new inhibitors. Arterioscler. Thromb. Vasc. Biol. 38, e28–e35 (2018).
Data availability 29. Moco, S., Martin, F. P. J. & Rezzi, S. Metabolomics view on gut microbiome
All data supporting the findings of this study are available at https://github.com/ modulation by polyphenol-rich foods. J. Proteome Res. 11, 4781–4790 (2012).
italodovalle/polyphenols and within the paper and its Supplementary 30. van Duynhoven, J. et al. Metabolic fate of polyphenols in the human
Information files. superorganism. Proc. Natl Acad. Sci. USA 108, 4531–4538 (2011).
31. Ottaviani, J. I., Heiss, C., Spencer, J. P. E., Kelm, M. & Schroeter, H.
Recommending flavanols and procyanidins for cardiovascular health:
Code availability revisited. Mol. Aspects Med. 61, 63–75 (2018).
Computer code is available at https://github.com/italodovalle/polyphenols.
32. Stalmach, A., Troufflard, S., Serafini, M. & Crozier, A. Absorption,
metabolism and excretion of Choladi green tea flavan-3-ols by humans.
Received: 27 August 2020; Accepted: 16 February 2021; Mol. Nutr. Food Res. 53, S44–53 (2009).
Published online: 19 March 2021 33. Meng, X. et al. Identification and characterization of methylated and
ring-fission metabolites of tea catechins formed in humans, mice, and rats.
References Chem. Res. Toxicol. 15, 1042–1050 (2002).
1. Khera, A. V. et al. Genetic risk, adherence to a healthy lifestyle, and coronary 34. Perez-Vizcaino, F., Duarte, J. & Santos-Buelga, C. The flavonoid paradox:
disease. N. Engl. J. Med. 375, 2349–2358 (2016). conjugation and deconjugation as key steps for the biological activity of
2. Arts, I. C. W. & Hollman, P. C. H. Polyphenols and disease risk in flavonoids. J. Sci. Food Agric. 92, 1822–1825 (2012).
epidemiologic studies. Am. J. Clin. Nutr. 81, 317S–325S (2005). 35. Shimoi, K. & Nakayama, T. Glucuronidase deconjugation in inflammation.
3. Wang, X., Ouyang, Y. Y., Liu, J. & Zhao, G. Flavonoid intake and risk of Methods Enzymol. 400, 263–272 (2005).
CVD: a systematic review and meta-analysis of prospective cohort studies. 36. Kaneko, A. et al. Glucuronides of phytoestrogen flavonoid enhance
Br. J. Nutr. 111, 1–11 (2014). macrophage function via conversion to aglycones by β-glucuronidase in
4. Neveu, V. et al. Phenol-Explorer: an online comprehensive database on macrophages. Immun. Inflamm. Dis. 5, 265–279 (2017).
polyphenol contents in foods. Database 2010, bap024 (2010). 37. Cheng, F., Kovács, I. A. & Barabási, A.-L. Network-based prediction of drug
5. Pérez-Jiménez, J., Neveu, V., Vos, F. & Scalbert, A. Systematic analysis of the combinations. Nat. Commun. 10, 1197 (2019).
content of 502 polyphenols in 452 foods and beverages: an application of the 38. Smalley, J. L., Gant, T. W. & Zhang, S.-D. Application of connectivity
phenol-explorer database. J. Agric. Food Chem. 58, 4959–4969 (2010). mapping in predictive toxicology based on gene-expression similarity.
6. Zhang, H. & Tsao, R. Dietary polyphenols, oxidative stress and antioxidant Toxicology 268, 143–146 (2010).
and anti-inflammatory effects. Curr. Opin. Food Sci. 8, 33–42 (2016). 39. Lamb, J. et al. The Connectivity Map: using gene-expression signatures
7. Boly, R. et al. Quercetin inhibits a large panel of kinases implicated in cancer to connect small molecules, genes, and disease. Science 313,
cell biology. Int. J. Oncol. 38, 833–842 (2011). 1929–1935 (2006).
8. Lacroix, S. et al. A computationally driven analysis of the polyphenol-protein 40. Amanzadeh, E. et al. Quercetin conjugated with superparamagnetic iron
interactome. Sci. Rep. 8, 2232 (2018). oxide nanoparticles improves learning and memory better than free quercetin
9. Hanhineva, K. et al. Impact of dietary polyphenols on carbohydrate via interacting with proteins involved in LTP. Sci. Rep. 9, 6876 (2019).
metabolism. Int. J. Mol. Sci. 11, 1365–1402 (2010). 41. Shaikh, J., Ankola, D. D., Beniwal, V., Singh, D. & Kumar, M. N. V. R.
10. Hervert-Hernández, D. & Goñi, I. Dietary polyphenols and human gut Nanoparticle encapsulation improves oral bioavailability of curcumin by at
microbiota: a review. Food Rev. Int. 27, 154–169 (2011). least 9-fold when compared to curcumin administered with piperine as
11. Zhang, S. et al. Dietary pomegranate extract and inulin affect gut microbiome absorption enhancer. Eur. J. Pharm. Sci. 37, 223–230 (2009).
differentially in mice fed an obesogenic diet. Anaerobe 48, 184–193 (2017). 42. Chao, E. C. & Henry, R. R. SGLT2 inhibition-A novel strategy for diabetes
12. Thazhath, S. S. et al. Administration of resveratrol for 5 wk has no effect on treatment. Nat. Rev. Drug Discov. 9, 551–559 (2010).
glucagon-like peptide 1 secretion, gastric emptying, or glycemic control in type 43. Caldera, M. et al. Mapping the perturbome network of cellular perturbations.
2 diabetes: a randomized controlled trial. Am. J. Clin. Nutr. 103, 66–70 (2016). Nat. Commun. 10, 5140 (2019).
154 Nature Food | VOL 2 | March 2021 | 143–155 | www.nature.com/natfoodYou can also read
